Article type
Year
Abstract
Background: Epidemiological studies may be subject to selective reporting biases, but empirical evidence thereof is limited.
Objectives: To probe into the extent of selective reporting in epidemiological studies.
Methods: We evaluated 389 epidemiological studies reporting at least one relative risk for a continuous risk factor in contrasts based on median, tertile, quartile or quintile categorizations. We evaluated the proportion and correlates of reporting statistically significant and non-statistically significant results and examined whether the magnitude of the presented relative risks (coined to be consistently >= 1.00) differs depending on the type of contrast used for the risk factor.
Results: Three hundred and forty-two studies (87.9%) reported at least one statistically significant relative risk in the abstract, while only 169 (43.4%) reported at least one non-statistically significant relative risk. Reporting of significant results was more common with structured abstracts, and less common in USA-based studies and those of cancer outcomes. The inverse associations were seen for reporting at least one statistically non-significant risk. Paradoxically, the smallest presented relative risks were based on extreme quintiles' contrasts; on average, the relative risk magnitude was 1.41-, 1.42-, and 1.36-fold larger in contrasts of extreme quartiles, extreme tertiles, and above vs. below median values, respectively (p<0.001).
Conclusions: Epidemiological investigations almost ubiquitously claim significant associations between risk factors and outcomes. For continuous risk factors, investigators present selectively contrasts between more extreme groups, when relative risks are inherently smaller. Selective reporting threatens the credibility of the epidemiological literature. Supported by a training PENED grant from the General Secretariat for Research and Technology, Greece.
Objectives: To probe into the extent of selective reporting in epidemiological studies.
Methods: We evaluated 389 epidemiological studies reporting at least one relative risk for a continuous risk factor in contrasts based on median, tertile, quartile or quintile categorizations. We evaluated the proportion and correlates of reporting statistically significant and non-statistically significant results and examined whether the magnitude of the presented relative risks (coined to be consistently >= 1.00) differs depending on the type of contrast used for the risk factor.
Results: Three hundred and forty-two studies (87.9%) reported at least one statistically significant relative risk in the abstract, while only 169 (43.4%) reported at least one non-statistically significant relative risk. Reporting of significant results was more common with structured abstracts, and less common in USA-based studies and those of cancer outcomes. The inverse associations were seen for reporting at least one statistically non-significant risk. Paradoxically, the smallest presented relative risks were based on extreme quintiles' contrasts; on average, the relative risk magnitude was 1.41-, 1.42-, and 1.36-fold larger in contrasts of extreme quartiles, extreme tertiles, and above vs. below median values, respectively (p<0.001).
Conclusions: Epidemiological investigations almost ubiquitously claim significant associations between risk factors and outcomes. For continuous risk factors, investigators present selectively contrasts between more extreme groups, when relative risks are inherently smaller. Selective reporting threatens the credibility of the epidemiological literature. Supported by a training PENED grant from the General Secretariat for Research and Technology, Greece.